Considering the situation that photovoltaicpower generation is affected by a variety of weather factors,a hybrid model was proposed based on warped Gaussian process to predict the power generation,where probability of photovoltaic power generation at any time in one day can be realized and prediction point and prediction interval can be obtained. Firstly,multivariate adaptive regression splines model was used to reduce multidimensional input variables,and to obtain the prior data of test. According to the type of weather,fuzzy C-means algorithm was then used to divide the training data and prior data of test,and to obtain the similar samples. The warped Gaussian process was also used to estimate the test data. Finally,bagging algorithm was used to realize the integrated study,and to obtain the prediction interval and prediction point. By the simulation and experimental results,the validity and reliability of this hybrid model was verified. The results show that the hybrid model improves both accuracy and practicability,compared with Gaussian process predictions and BP quantile regression neural network predictions.